Kybernetika 48 no. 4, 809-823, 2012

A note on prediction for discrete time series

This article was granted Editor's award of the year 2012Editor's award 2012

Gusztáv Morvai and Benjamin Weiss


Let $\{X_n\}$ be a stationary and ergodic time series taking values from a finite or countably infinite set ${\cal X}$ and that $f(X)$ is a function of the process with finite second moment. Assume that the distribution of the process is otherwise unknown. We construct a sequence of stopping times $\lambda_n$ along which we will be able to estimate the conditional expectation $E(f(X_{\lambda_n+1})|X_0,\dots,X_{\lambda_n} )$ from the observations $(X_0,\dots,X_{\lambda_n})$ in a point wise consistent way for a restricted class of stationary and ergodic finite or countably infinite alphabet time series which includes among others all stationary and ergodic finitarily Markovian processes. If the stationary and ergodic process turns out to be finitarily Markovian (in particular, all stationary and ergodic Markov chains are included in this class) then $ \lim_{n\to \infty} \frac{n}{\lambda_n}>0$ almost surely. If the stationary and ergodic process turns out to possess finite entropy rate then $\lambda_n$ is upper bounded by a polynomial, eventually almost surely.


stationary processes, nonparametric estimation


62G05, 60G25, 60G10


  1. D. H. Bailey: Sequential Schemes for Classifying and Predicting Ergodic Processes. Ph.D. Thesis, Stanford University 1976.   CrossRef
  2. A. Berlinet, I. Vajda and E. C. van der Meulen: About the asymptotic accuracy of Barron density estimates. IEEE Trans. Inform. Theory 44 (1998), 3, 999-1009.   CrossRef
  3. K. L. Chung: A note on the ergodic theorem of information theory. Ann. Math. Statist. 32 (1961), 612-614.   CrossRef
  4. T. M. Cover and J. Thomas: Elements of Information Theory. Wiley, 1991.   CrossRef
  5. I. Csiszár and P. Shields: The consistency of the BIC Markov order estimator. Ann. Statist. 28 (2000), 1601-1619.   CrossRef
  6. I. Csiszár: Large-scale typicality of Markov sample paths and consistency of MDL order estimators. IEEE Trans. Inform. Theory 48 (2002), 1616-1628.   CrossRef
  7. G. A. Darbellay and I. Vajda: Estimation of the information by an adaptive partitioning of the observation space. {IEEE Trans. Inform. Theory 45 (1999), 4, 1315-1321.}   CrossRef
  8. J. Feistauerová and I. Vajda: Testing system entropy and prediction error probability. IEEE Trans. Systems Man Cybernet. 23 (1993), 5 1352-1358.   CrossRef
  9. L. Györfi, G. Morvai and S. Yakowitz: Limits to consistent on-line forecasting for ergodic time series. {IEEE Trans. Inform. Theory} 44 (1998), 886-892.   CrossRef
  10. L. Györfi, G. Morvai and I. Vajda: Information-theoretic methods in testing the goodness of fit. {In: Proc. 2000 IEEE Internat. Symposium on Information Theory}, ISIT 2000, New York and Sorrento, p. 28.   CrossRef
  11. W. Hoeffding: Probability inequalities for sums of bounded random variables. {J. Amer. Statist. Assoc.} 58 (1963), 13-30.   CrossRef
  12. S. Kalikow: Random Markov processes and uniform martingales. {Israel J. Math.} 71 (1990), 33-54.   CrossRef
  13. M. Keane: Strongly mixing g-measures. {Invent. Math. } 16 (1972), 309-324.   CrossRef
  14. H. Luschgy, L. A. Rukhin and I. Vajda: Adaptive tests for stochastic processes in the ergodic case. {Stochastic Process. Appl.} 45 (1993), 1, 45-59.   CrossRef
  15. G. Morvai and I. Vajda: A survay on log-optimum portfolio selection. In: Second European Congress on Systems Science, Afcet, Paris 1993, pp. 936-944.   CrossRef
  16. G. Morvai and B. Weiss: Prediction for discrete time series. {Probab. Theory Related Fields} 132 (2005), 1-12.   CrossRef
  17. G. Morvai and B. Weiss: Estimating the memory for finitarily Markovian processes. { Ann. Inst. H. Poincaré Probab. Statist.} 43 (2007), 15-30.   CrossRef
  18. B. Ya. Ryabko: Prediction of random sequences and universal coding. {Problems Inform. Transmission} 24 (1988), 87-96.   CrossRef
  19. B. Ryabko: Compression-based methods for nonparametric prediction and estimation of some characteristics of time series. {IEEE Trans. Inform. Theory} 55 (2009), 9, 4309-4315.   CrossRef
  20. I. Vajda and F. Österreicher: Existence, uniqueness and evaluation of log-optimal investment portfolio. {Kybernetika} 29 (1993), 2, 105-120.   CrossRef
  21. I. Vajda and P. Harremoës: On the Bahadur-efficient testing of uniformity by means of entropy. IEEE Trans. Inform. Theory 54 (2008), 321-331.   CrossRef